Repository navigation
[Windows Desktop][Codex Cloud] Legacy environment migration stuck: Couldn't save environment / Couldn't publish #49568
Description
What version of the Codex App are you using (From “About Codex” dialog)?
Installed Windows package: OpenAI.Codex 26.928.2636.0 (read from Get-AppxPackage on September 30, 2026). The About dialog version was not separately captured.
What subscription do you have?
Not captured.
What platform is your computer?
Windows, desktop app. The environment being migrated runs in Codex Cloud.
What issue are you seeing?
Migrating an existing Legacy Codex Cloud environment into the current Codex Cloud environment workflow is stuck. Onboarding can update the configuration draft, but the environment cannot be saved/published and setup cannot complete.
The environment panel displays these exact errors:
Couldn't publish
Couldn't save changes
Some changes couldn't be saved. Review the details and try again.
Couldn't save environment
Needs attention
An earlier settings view had disabled editing controls. The latest environment panel shows internet access switched on and All (unrestricted), while the setup attempt reports disabled network access and an unreachable proxy.
The setup chats recorded the runtime's /etc/codex/network-policy.json with empty HTTP egress rules. A dependency registry request failed:
curl: (7) Failed to connect to proxy port 8080 after 0 ms: Could not connect to server
Two onboarding attempts encountered the same blocker. Installation scripts, startup instructions, and a dependency domain allowlist were saved through the onboarding configuration-draft tools, but successful draft updates did not produce a usable, published environment. Dependency installation, Bun checks, application startup, and build validation remain incomplete.
What steps can reproduce the bug?
Observed sequence on the affected account (not yet reproduced on another account):
- Start migrating an existing legacy Codex Cloud repository environment to the current Codex Cloud workflow.
- Run the suggested Cloud Environment Onboarding: Setup task.
- Let onboarding inspect the checkout and update the install script, start skill, and package/download domain allowlist.
- Review the environment settings and attempt to save/publish.
- Observe Couldn't save environment / Couldn't publish.
- Retry onboarding in a second setup chat. It again reports disabled networking and an unreachable
proxy:8080, leaving setup unfinished.
What is the expected behavior?
The migrated configuration should save and publish successfully, and the runtime network policy should reflect the applied settings so onboarding can install dependencies and validate startup.
If migration cannot complete, please provide a supported way to repair/reset the migration or recreate the environment while preserving its setup configuration.
Additional information
Feedback was successfully submitted through the desktop app. The confirmation dialog instructed the user to open a GitHub issue if the problem persists.
Feedback ID: 01a0f1cc-d9e0-721d-ade3-6d77bf5fa026
Please correlate that feedback/session with the failed environment-save/publish request and provisioning/network-policy logs.
The root cause is unknown. The evidence establishes a failed save/publish, an earlier configuration/runtime mismatch, and failed proxy connectivity; it does not establish which backend operation caused the failure. No failed save HTTP status or backend response body was available.
Error text above is transcribed from user-provided screenshots and setup-chat output. Private repository names, local user paths, configuration identifiers, credentials, and full transcripts are omitted from this public report.
Activity
ElierHG commentedon Sep 30, 2026
Screenshot showing the failed environment save/publish, with internet access enabled in the environment panel:
sashasochka commentedon Sep 30, 2026
Same error, also on windows / latest
MindTooth commentedon Oct 2, 2026
Same on macOS (
26.930.21537) and web as of October 2nd.papadeeley commentedon Oct 4, 2026
Additional report from macOS for the same environment save failure and proxy failure.
OpenAI Support asked me to file the affected environment and session details on GitHub on 4 October 2026 (case 16399983). I am adding them here because this open issue matches the failures in my setup chat.
What issue are you seeing?
Cloud environment setup for
papadeeley/drivetrackcannot complete. The setup chat records successful configuration-draft updates, followed by a settings save failure. The running environment still reported network access disabled after the web allowlist was saved in the draft.The setup chat contains this exact command and result:
A GitHub read failed with the same proxy connection error. Dependency installation and application startup, tests, and build validation could not complete.
What steps can reproduce the bug?
Observed sequence in the affected session on 30 September 2026; this has not been reproduced on a separate account:
papadeeley/drivetrack, branchmain.The draft allowlist contained
registry.npmjs.org,nodejs.org,fonts.googleapis.com, andfonts.gstatic.com.What is the expected behavior?
Environment settings should save and publish. The running environment should then use the applied network settings so setup can install dependencies and verify the app. If the environment is stuck, please provide a way to repair, reset, or delete it.
Additional information
papadeeley/drivetrack01a0f142-6ee2-753c-8b23-9c6744f3a04ceb6720d0-4799-4f7b-a8ca-28b48cd22b8b~asenvcfg_cee4e1b9955081919e08f56d6a922052The evidence above comes from the saved setup chat dated 30 September. On 4 October, I read that chat again to prepare this report. I could not check the current settings or take new screenshots because Computer Use blocked access to the Codex app. The root cause, current control states, and failed save HTTP response remain unknown. Please use the session and configuration IDs to locate the failed save and provisioning logs.